Diffusion-weighted magnetic resonance imagingIn cervical cancer : beyond apparent diffusion coefficient

Abstract

Diffusion-weighted magnetic resonance imaging (DW-MRI) tracks the biological changes in the tumour microenvironment. Intravoxel incoherent imaging (IVIM), an extension of DW-MRI, can simultaneously explore the cellular and microcirculation in the cervical cancer microenvironment without the dependence on exogenous contrast agent. IVIM describes the signal decay as bi-exponential; attributed to the incoherent motion of blood within the unorganised capillaries described by f (perfusion fraction) and D*(pseudo-diffusion coefficient), and the true diffusion of water molecules by D (true diffusion coefficient). The series of studies presented in this thesis aimed to explore the feasibility of implementing IVIM and its potential clinical value in cervical cancer, with a secondary aim of discussing other emerging analyses of DW-MRI and potential clinical applications in cervical cancer. Treatment-naïve cervical cancer patients (n=74) were prospectively recruited to undergo MRI with DW-MRI (13 b-values). Among them, 25 patients underwent additional dynamic contrast-enhanced MRI (DCE-MRI), 39 patients had sequential MRI and 29 patients had 18F-fluoro-deoxyglucose positron emission tomography/computed tomography (FDG-PET/CT). Analyses of DW-MRI included IVIM, generation of the apparent diffusion coefficient (ADC) based on the mono-exponential signal decay model, texture analysis based on gray-level matrices (GLM) and tumour segmentation using k-means clustering method. We showed that cervical cancer exhibited bi-exponential IVIM effect (f 14.9±2.6%) with unique IVIM signatures when compared to normal and benign tissues in the female pelvis. There were moderate correlations between IVIM-derived perfusion parameters and those from DCE-MRI, ranging from r=0.42-0.58. The correlations were only moderate, suggesting that both provided complementary information on the tumour microcirculation in cervical cancer. Based on sequential MRI before and after chemoradiation, the changes in D of the irradiated marrow were associated with the occurrence of haematological toxicity. GLM of ADC, which relied on the distribution of the gray values and the relationships with neighbouring voxels, was associated with tumour grading in cervical cancer while the histogram skewness and kurtosis were higher in patients with nodal metastases. However, tumour segmentation was more precise with FDG-PET/CT based on the threshold of 30% of the maximum standardised uptake value (MTV30) than DW-MRI k-means clustering tumour segmentation method, potentially allowing for dose modulation to subvolumes of cervical cancer with more aggressive features. In summary, these early studies show that DW-MRI has promising clinical application in cervical cancer beyond the simple quantification of ADC. Results from IVIM are encouraging covering different aspects of the clinical management, from tumour identification to treatment monitoring. Data on texture analysis in cervical cancer is limited but holds potential with our preliminary results. Nevertheless, there are challenges in harmonisation of data acquisition and analysis to reduce variability within and between centres, so that these early results could be prospectively tested in multicentre studies and evaluated for their clinical impacts before clinical translation.